CN103108187B - The coded method of a kind of 3 D video, coding/decoding method, encoder - Google Patents

Abstract

The present invention proposes the coded method of a kind of 3 D video, coding/decoding method and encoder thereof, wherein coded method includes step: depth map and texture maps to reference view encode, and obtains reference view coding and rebuilding depth map and reference view coding and rebuilding texture maps；According to reference view coding and rebuilding depth map and reference view coding and rebuilding texture maps and corresponding camera parameter, by three-dimension intensity, obtain target view synthesized reference frame；Obtain reference view original texture figure and as primary signal, using target view synthesized reference frame as adding the signal after making an uproar, carry out Wiener filtering, obtain optimization aim View Synthesis reference frame, and solve wiener filter coefficients；Optimization aim View Synthesis reference frame is added set of reference frames, and wiener filter coefficients is write code stream.The present invention has raising code efficiency, the advantage improving video quality.

Description

The coded method of a kind of 3 D video, coding/decoding method, encoder

Technical field

The present invention relates to 3 d video encoding technical field, particularly propose a kind of 3 d video encoding/solution based on Wiener filtering Code method and coding/decoding device.

Background technology

Along with the development of multimedia communication technology, traditional two-dimensional image video, the even 3-D view of fixed view Video, can not meet the visually-perceptible demand of people again.In recent years, in numerous applications such as medical science, military affairs, amusements The most all occur in that the demand for free viewpoint video and 3 D video.For instance, it is possible to freely switch the free view-point at viewing visual angle Display device, and the three-dimensional television of different wide viewing angle video is shown to the beholder of diverse location.In order to realize these application, Efficient multiple view video coding technology is particularly important.

In multiple view video coding, due to the difference of shooting angle, multiple video cameras when shooting Same Scene, regarding of generation Certain geometric distortion is there is between point.View Synthesis prediction (VSP) technology proposes for compensating geometric distortion, and it is main Thought is: at texture video information one visual point image of synthesis of coding side recycling depth information and coding and rebuilding, and will It is used as the reference picture of current encoded image, and this algorithm makes the visual point image generated be more nearly than interview reference image Current encoded image, such that it is able to the data redundancy greatly reduced between viewpoint.Saying further, View Synthesis Predicting Technique is at figure As the other realization of frame level can be briefly described for: utilize the geological information of reference view and scene to synthesize the image of virtual view, And using the composograph of these virtual views as present encoding viewpoint reference frame be used for predictive coding.Therefore, View Synthesis The picture quality of reference frame largely effects on precision and the accuracy of coding prediction.If the picture quality of View Synthesis reference frame can be improved, Raising coding efficiency that then can be certain.

The View Synthesis Predicting Technique of prior art disadvantageously, the quality of View Synthesis reference frame is not high enough, thus directly shadow Ring the coding efficiency of relevant video sequence.The present invention is by the method utilizing Wiener filtering, the View Synthesis reference that will have generated Frame is filtered optimizing, and improves the quality of View Synthesis reference frame, and then improves precision and the coding efficiency of coding.

Summary of the invention

It is contemplated that solve one of above-mentioned technical problem the most to a certain extent or provide at a kind of useful business selection. To this end, it is an object of the present invention to propose 3 d video encoding/coding/decoding method that a kind of code efficiency is high, video quality is good. Further object is that and propose 3 d video encoding/decoding apparatus that a kind of code efficiency is high, video quality is good.

The coded method of 3 D video according to embodiments of the present invention, including: the coded method of a kind of 3 D video, its feature exists In, including: depth map and the texture maps of reference view are encoded by S1., obtain reference view coding and rebuilding depth map D_rec With reference view coding and rebuilding texture maps T_rec；S2. according to described reference view coding and rebuilding depth map D_rec and reference view Coding and rebuilding texture maps T_rec and corresponding camera parameter, by three-dimension intensity, obtain target view synthesized reference frame VS_rec；S3. reference view original texture figure T_orig is obtained and as primary signal, by described target view synthesized reference frame VS_rec, as adding the signal after making an uproar, carries out Wiener filtering, obtains optimization aim View Synthesis reference frame VS_rec_wiener, And solve wiener filter coefficients；And described optimization aim View Synthesis reference frame VS_rec_wiener is added reference by S4. Frame collection, and described wiener filter coefficients is write code stream.

Alternatively, by solving the Wiener Hopf equation described wiener filter coefficients of calculating.

Alternatively, in described Wiener filter: definition input pixel x_kWith Wiener filter output pixel z_k, wherein wiener filter The output z of ripple device_kBy wave filter support { reconstruction pixel y in S}_iComposition, support size is L+1, and weights are c_i, then wiener Filter function is:Input pixel x_kWith pixel z after Wiener filtering_kBetween residual signals C be defined as: error_k=z_k-x_k, there is filter tap { c by making_iMean square deviation minimize and optimize Wiener filter: In order to findMinima, to c_iDifferentiation also derives filter tap by making derivative be equal to zero: $\frac{\∂}{\∂c_i}E\left[{\mathrm{error}}_k^2\right]=2\left(\underset{j\∈\left\{S\right\}}{\mathrm{\Σ}}E\right\{\left(y_i\right)\left(y_j\right)\left\}{c}_j\right)-2E\left[\left(y_i\right)\left(x_k\right)\right]=0,$ Wherein i=0 ..., L, note { auto-correlation function of y} and { y} and { x} Cross-correlation function be respectively r_yy(i)=E[y_k y_k+1] and r_xy(i)=E[x_k y_k+1], then it is rewritten as with matrix form:

Thus Wiener filtering coefficient C} can be derived as with matrix form:

The coding/decoding method of 3 D video according to embodiments of the present invention, it is characterised in that including: S1. receives and is wanted by according to right The code stream asking the coded method of the 3 D video described in any one of 1-3 to obtain, prepares Decoded Reference two field picture frame by frame；S2. institute is judged State the type of reference frame image, if target view synthesized reference frame, then perform S31-S34, if independent view reference Frame, then perform S4；S31. from code stream, extract reference view coding and rebuilding depth map D_rec and reference view coding and rebuilding texture Figure T_rec；S32. from code stream, extract corresponding camera parameter, in conjunction with described reference view coding and rebuilding depth map D_rec and Reference view coding and rebuilding texture maps T_rec, by three-dimension intensity, obtains target view synthesized reference frame VS_rec；S33. From code stream, extract described Wiener filtering coefficient, described target view synthesized reference frame VS_rec is carried out noise reduction filtering, obtains Final goal View Synthesis reference frame VS_rec_final；Final goal View Synthesis reference frame VS_rec_final is read with S34. Information, complete the decoding process of video image；S4. directly read the information of described reference frame image, complete video image Decoding process.

The encoder of 3 D video according to embodiments of the present invention, including: S1. coding and rebuilding module, described reconstruction module is used for Depth map and texture maps to reference view encode, and obtain reference view coding and rebuilding depth map D_rec and reference view is compiled Code rebuilds texture maps T_rec；S2. geometric transformation module, is that set transform module is for deep according to described reference view coding and rebuilding Degree schemes D_rec and reference view coding and rebuilding texture maps T_rec and corresponding camera parameter, by three-dimension intensity, To target view synthesized reference frame VS_rec；S3. Wiener filtering computing module, described filtration module is with reference view original texture Figure T_orig, as primary signal, using described target view synthesized reference frame VS_rec as adding the signal after making an uproar, carries out wiener Filtering, obtains optimization aim View Synthesis reference frame VS_rec_wiener, and solves wiener filter coefficients；And S4. code stream Sending module, described optimization aim View Synthesis reference frame VS_rec_wiener is added set of reference frames by described coding sending module, And described wiener filter coefficients is write code stream, send subsequently.

Alternatively, in described Wiener filtering computing module, calculate described wiener filter coefficients by solving Wiener Hopf equation.

Alternatively, in described Wiener filter: definition input pixel x_kWith Wiener filter output pixel z_k, wherein Wiener filtering The output z of device_kBy wave filter support { reconstruction pixel y in S}_iComposition, support size is L+1, and weights are c_i, then wiener filter Ripple device function is:Input pixel x_kWith pixel z after Wiener filtering_kBetween residual signals C be defined as: error_k=z_k-x_k, there is filter tap { c by making_iMean square deviation minimize and optimize Wiener filter: In order to findMinima, to c_iDifferentiation also derives filter tap by making derivative be equal to zero: $\frac{\∂}{\∂c_i}E\left[{\mathrm{error}}_k^2\right]=2\left(\underset{j\∈\left\{S\right\}}{\mathrm{\Σ}}E\right\{\left(y_i\right)\left(y_j\right)\left\}{c}_j\right)-2E\left[\left(y_i\right)\left(x_k\right)\right]=0,$ Wherein i=0 ..., L, note { auto-correlation function of y} and { y} and { x} Cross-correlation function be respectively r_yy(i)=E[y_k y_k+1] and r_xy(i)=E[x_k y_k+1], then it is rewritten as with matrix form:

Thus Wiener filtering coefficient C} can be derived as with matrix form:

The decoder of 3 D video according to embodiments of the present invention, including: code stream receiver module, described code stream receiver module is used for Receive by the encoder transmitted stream according to the 3 D video described in any one of claim 5-7, prepare decoded reference frame frame by frame Image；Judge module, described judge module is for judging the type of described reference frame image, if target view synthesized reference Frame, then sequentially enter reconstruction module, geometric transformation module, Wiener filtering computing module and decoding read module, if independent View reference frame, then be directly entered described decoding read module；Described reconstruction module, described reconstruction module is for carrying from code stream Take reference view coding and rebuilding depth map D_rec and reference view coding and rebuilding texture maps T_rec；Described geometric transformation module, Described geometric transformation module is for extracting corresponding camera parameter, in conjunction with described reference view coding and rebuilding depth map from code stream D_rec and reference view coding and rebuilding texture maps T_rec, by three-dimension intensity, obtain target view synthesized reference frame VS_rec；Described Wiener filtering computing module, described Wiener filtering computing module is for extracting described Wiener filtering system from code stream Number, carries out noise reduction filtering to described target view synthesized reference frame VS_rec, obtains final goal View Synthesis reference frame VS_rec_final, and it is sent to described decoding read module as the reference frame image after updating；Described decoding read module, For reading the information of described reference frame image, complete the decoding process of video image.

In the present invention, View Synthesis reference frame generates in video encoding-decoding process, owing to using enhancing based on Wiener filtering Visual point synthesizing method, is filtered View Synthesis reference frame, thus is had good inhibiting effect to noise, as volume Reference frame during decoding, can improve the accuracy of estimation, reduces forecast error, improves code efficiency, change simultaneously The subjective quality of video is rebuild after being apt to compression coding.Additionally, the View Synthesis reference frame in the present invention can be in video decoders Rebuild and generate and dynamically update, from the video data the most extra without transmission, as long as transmission is corresponding in video present frame The Wiener filtering coefficient of View Synthesis reference frame, saved resource.

The additional aspect of the present invention and advantage will part be given in the following description, and part will become apparent from the description below, Or recognized by the practice of the present invention.

Accompanying drawing explanation

Above-mentioned and/or the additional aspect of the present invention and advantage the accompanying drawings below description to embodiment will be apparent from from combining and Easy to understand, wherein:

Fig. 1 is the flow chart of the coded method of 3 D video according to embodiments of the present invention；

Fig. 2 is the flow chart of the coding/decoding method of 3 D video according to embodiments of the present invention；

Fig. 3 is the structure chart of the encoder of 3 D video according to embodiments of the present invention；

Fig. 4 is the structure chart of the decoder of 3 D video according to embodiments of the present invention.

Detailed description of the invention

Embodiments of the invention are described below in detail, and the example of described embodiment is shown in the drawings, the most identical or Similar label represents same or similar element or has the element of same or like function.Describe below with reference to accompanying drawing Embodiment is exemplary, it is intended to is used for explaining the present invention, and is not considered as limiting the invention.

In describing the invention, it is to be understood that term " " center ", " longitudinally ", " laterally ", " length ", " width ", " thickness Degree ", " on ", D score, "front", "rear", "left", "right", " vertically ", " level ", " top ", " end " " interior ", " outward ", " suitable Hour hands ", the orientation of the instruction such as " counterclockwise " or position relationship be based on orientation shown in the drawings or position relationship, merely to just Describe in the description present invention and simplification rather than indicate or imply that the device of indication or element must have specific orientation, Yi Te Fixed azimuth configuration and operation, be therefore not considered as limiting the invention.

Additionally, term " first ", " second " are only used for describing purpose, and it is not intended that instruction or hint relative importance or The implicit quantity indicating indicated technical characteristic.Thus, define " first ", the feature of " second " can be expressed or implicit Ground includes one or more this feature.In describing the invention, " multiple " are meant that two or more, unless Separately there is the most concrete restriction.

In the present invention, unless otherwise clearly defined and limited, term " install ", " being connected ", " connection ", the art such as " fixing " Language should be interpreted broadly, and connects for example, it may be fixing, it is also possible to be to removably connect, or be integrally connected；It can be machine Tool connects, it is also possible to be electrical connection；Can be to be joined directly together, it is also possible to be indirectly connected to by intermediary, can be two units Connection within part.For the ordinary skill in the art, can understand that above-mentioned term is in the present invention as the case may be In concrete meaning.

In the present invention, unless otherwise clearly defined and limited, fisrt feature second feature it " on " or D score permissible Directly contact including the first and second features, it is also possible to include that the first and second features are not directly contact but by them Between other characterisation contact.And, fisrt feature second feature " on ", " top " and " above " include first Feature is directly over second feature and oblique upper, or is merely representative of fisrt feature level height higher than second feature.First is special Levy second feature " under ", " lower section " and " below " include fisrt feature immediately below second feature and obliquely downward, or only Only represent that fisrt feature level height is less than second feature.

For making those skilled in the art be more fully understood that the present invention, the existing framework to View Synthesis Predicting Technique is described further.

In the coding stage of View Synthesis Predicting Technique, encoder is based on encoded depth image associated and texture image, profit With three-dimension intensity, generate the View Synthesis reference frame of current encoded image.Owing to View Synthesis reference frame also may be used in decoding end Based on pertinent image information, by three-dimension intensity, reproduction generates.So, based on regarding that View Synthesis Predicting Technique generates Point synthesized reference frame is not required to coding write code stream, thus greatly reduces encoding code stream, improves code efficiency.If it is current Be coded frame place viewpoint be absolute coding viewpoint, then there is not View Synthesis reference frame in it；If being currently coded frame place viewpoint For forward predictive coded viewpoint, then its View Synthesis reference frame adjacent viewpoint in the same time is encoded texture maps and depth map are carried out View Synthesis；If be currently coded frame place viewpoint be bi-directional predictive coding viewpoint, then its View Synthesis reference frame is by the same time two Texture maps and depth map that individual adjacent viewpoint is encoded carry out View Synthesis, then are weighted averagely merging by two picture frames, If the pixel value P of forward prediction viewpoint dummy synthesis image_fRepresenting, the pixel value of back forecast viewpoint dummy synthesis image is used P_bRepresent, then the pixel value P of the reference frame of final View Synthesis_refIt is represented by, P_ref=(1-α)P_f+αP_b, 0 < α < 1(α according to Distance between viewpoint and different.Distance is the nearest, and the value of α is the biggest).As can be seen here, the picture quality of the reference frame of View Synthesis is big Affect greatly precision and the accuracy of coding prediction.If the picture quality of View Synthesis reference frame can be improved, then raising that can be certain Coding efficiency.

The present invention utilizes encoded image information to be predicted encoding to the picture frame of follow-up coding as reference frame, wherein reference Frame comprises View Synthesis reference frame based on three-dimension intensity.This View Synthesis reference frame is encoded by adjacent viewpoint in the same time Texture maps and depth map carry out View Synthesis.Then, use Wiener filter to be filtered the View Synthesis reference frame generated, Improve the quality of view.View Synthesis reference frame after renewal is predicted when follow-up picture frame encodes.Correspondingly, The decoding process of the present invention has also used same principle to improve code efficiency and to improve video quality.

As it is shown in figure 1, be the flow chart of the coded method of the 3 D video according to the embodiment of the present invention, including: a kind of three-dimensional regards The coded method of frequency, it is characterised in that including: depth map and the texture maps of reference view are encoded by S1., obtain reference Viewpoint coding and rebuilding depth map D_rec and reference view coding and rebuilding texture maps T_rec；S2. deep according to reference view coding and rebuilding Degree schemes D_rec and reference view coding and rebuilding texture maps T_rec and corresponding camera parameter, by three-dimension intensity, To target view synthesized reference frame VS_rec；S3. reference view original texture figure T_orig is obtained and as primary signal, by mesh Mark View Synthesis reference frame VS_rec, as adding the signal after making an uproar, carries out Wiener filtering, obtains optimization aim View Synthesis reference Frame VS_rec_wiener, and solve wiener filter coefficients；And S4. is by optimization aim View Synthesis reference frame VS_rec_wiener adds set of reference frames, and wiener filter coefficients is write code stream.

Alternatively, by solving the Wiener Hopf equation described wiener filter coefficients of calculating.

Alternatively, in described Wiener filter: definition input pixel x_kWith Wiener filter output pixel z_k, wherein wiener filter The output z of ripple device_kBy wave filter support { reconstruction pixel y in S}_iComposition, support size is L+1, and weights are c_i, then wiener Filter function is:Input pixel x_kWith pixel z after Wiener filtering_kBetween residual signals C be defined as: error_k=z_k-x_k, there is filter tap { c by making_iMean square deviation minimize and optimize Wiener filter: In order to findMinima, to c_iDifferentiation also derives filter tap by making derivative be equal to zero: $\frac{\&PartialD;}{\&PartialD;c_i}E\left[{\mathrm{error}}_k^2\right]=2\left(\underset{j\&Element;\left\{S\right\}}{\mathrm{\&Sigma;}}E\right\{\left(y_i\right)\left(y_j\right)\left\}{c}_j\right)-2E\left[\left(y_i\right)\left(x_k\right)\right]=0,$ Wherein i=0 ..., L, note { auto-correlation function of y} and { y} and { x} Cross-correlation function be respectively r_yy(i)=E[y_k y_k+1] and r_xy(i)=E[x_k y_k+1], then it is rewritten as with matrix form:

Thus Wiener filtering coefficient C} can be derived as with matrix form:

As in figure 2 it is shown, be the flow chart of the coding/decoding method of the 3 D video according to the embodiment of the present invention, it is characterised in that including: S1. receive by the code stream obtained according to the coded method of the 3 D video described in any one of claim 1-3, prepare to decode ginseng frame by frame Examine two field picture；S2. judge the type of reference frame image, if target view synthesized reference frame, then perform S31-S34, if It is independent view reference frame, then performs S4；S31. from code stream, reference view coding and rebuilding depth map D_rec is extracted and with reference to regarding Point coding and rebuilding texture maps T_rec；S32. from code stream, corresponding camera parameter is extracted, in conjunction with the reference view coding and rebuilding degree of depth Figure D_rec and reference view coding and rebuilding texture maps T_rec, by three-dimension intensity, obtain target view synthesized reference frame VS_rec；S33. from code stream, extract Wiener filtering coefficient, target view synthesized reference frame VS_rec carried out noise reduction filtering, Obtain final goal View Synthesis reference frame VS_rec_final；Final goal View Synthesis reference frame is read with S34. The information of VS_rec_final, completes the decoding process of video image；S4. directly read the information of reference frame image, complete video The decoding process of image.

As it is shown on figure 3, be the structure chart of the encoder 1000 of the 3 D video according to the embodiment of the present invention, including: S1. encodes Rebuild module 1100, for depth map and the texture maps of reference view are encoded, obtain the reference view coding and rebuilding degree of depth Figure D_rec and reference view coding and rebuilding texture maps T_rec；S2. geometric transformation module 1200, for encoding according to reference view Rebuild depth map D_rec and reference view coding and rebuilding texture maps T_rec and corresponding camera parameter, become by three-dimensional geometry Change, obtain target view synthesized reference frame VS_rec；S3. Wiener filtering computing module 1300, with reference view original texture figure T_orig, as primary signal, using target view synthesized reference frame VS_rec as adding the signal after making an uproar, carries out Wiener filtering, Obtain optimization aim View Synthesis reference frame VS_rec_wiener, and solve wiener filter coefficients；And S4. code stream sends mould Block 1400, adds set of reference frames by optimization aim View Synthesis reference frame VS_rec_wiener, and by described Wiener filter system Number write code stream, sends subsequently.

Alternatively, in described Wiener filtering computing module 1300, calculate described Wiener filter by solving Wiener Hopf equation Coefficient.

Alternatively, in described Wiener filter: definition input pixel x_kWith Wiener filter output pixel z_k, wherein Wiener filtering The output z of device_kBy wave filter support { reconstruction pixel y in S}_iComposition, support size is L+1, and weights are c_i, then wiener filter Ripple device function is:Input pixel x_kWith pixel z after Wiener filtering_kBetween residual signals C be defined as: error_k=z_k-x_k, there is filter tap { c by making_iMean square deviation minimize and optimize Wiener filter: In order to findMinima, to c_iDifferentiation also derives filter tap by making derivative be equal to zero: $\frac{\&PartialD;}{\&PartialD;c_i}E\left[{\mathrm{error}}_k^2\right]=2\left(\underset{j\&Element;\left\{S\right\}}{\mathrm{\&Sigma;}}E\right\{\left(y_i\right)\left(y_j\right)\left\}{c}_j\right)-2E\left[\left(y_i\right)\left(x_k\right)\right]=0,$ Wherein i=0 ..., L, note { auto-correlation function of y} and { y} and { x} Cross-correlation function be respectively r_yy(i)=E[y_k y_k+1] and r_xy(i)=E[x_k y_k+1], then it is rewritten as with matrix form:

Thus Wiener filtering coefficient C} can be derived as with matrix form:

As shown in Figure 4, for the structure chart of the decoder 2000 of the 3 D video according to the embodiment of the present invention, including: code stream connects Receive module 2100, for receiving by encoder 1000 transmitted stream of the 3 D video according to any one of claim 5-7, Prepare Decoded Reference two field picture frame by frame；Judge module 2200, for judging the type of reference frame image, if target view Synthesized reference frame, then sequentially enter reconstruction module 2300, geometric transformation module 2400, Wiener filtering computing module 2500 are conciliate Code read module 2600, if independent view reference frame, is then directly entered described decoding read module 2600；Rebuild module 2300, for extracting reference view coding and rebuilding depth map D_rec and reference view coding and rebuilding texture maps T_rec from code stream； Geometric transformation module 2400, for extracting corresponding camera parameter, in conjunction with reference view coding and rebuilding depth map from code stream D_rec and reference view coding and rebuilding texture maps T_rec, by three-dimension intensity, obtain target view synthesized reference frame VS_rec；Wiener filtering computing module 2500, for extracting Wiener filtering coefficient, to target view synthesized reference from code stream Frame VS_rec carries out noise reduction filtering, obtains final goal View Synthesis reference frame VS_rec_final, and is sent to decoding reading mould Block 2600 is as the reference frame image after updating；Decoding read module 2600, for reading the information of described reference frame image, Complete the decoding process of video image.

In the present invention, View Synthesis reference frame generates in video encoding-decoding process, owing to using enhancing based on Wiener filtering Visual point synthesizing method, is filtered View Synthesis reference frame, thus is had good inhibiting effect to noise, as volume Reference frame during decoding, can improve the accuracy of estimation, reduces forecast error, improves code efficiency, change simultaneously The subjective quality of video is rebuild after being apt to compression coding.Additionally, the View Synthesis reference frame in the present invention can be in the solution of 3 D video Code device 2000 is rebuild and generates and dynamically update, from the video data the most extra without transmission, if current at video Frame transmits the Wiener filtering coefficient of corresponding View Synthesis reference frame.Although it should be noted that obtaining excellent in coding stage Change target view synthesized reference frame VS_rec_winer, but the decoding stage does not use it, but only with transmitting a wiener filter Reference frame used when wave system number just can go back original encoding, thus saved resource.

Without loss of generality, in order to make technical staff be more fully understood that the present invention, applicant is existing as a example by two-way stereo scopic video coding It is described.The explanation of technical scheme that the embodiment of the present invention provide is given below.The enforcement step of specific coding/decoding is as follows Shown in:

1. depth map and the texture maps of reference view are encoded by the coding side at three-dimensional video system, and obtain coding and rebuilding Depth map D_rec and texture maps T_rec.

2., according to depth map D_rec, texture maps T_rec and corresponding camera parameter, by three-dimension intensity, obtain target The View Synthesis image VS_rec of viewpoint.

3. the texture maps without compression taking target view the most in the same time is T_orig.

4., using T_orig as primary signal, VS_rec is to add the signal after making an uproar.VS_rec is carried out Wiener filtering, obtains VS_rec_wiener, and it is calculated corresponding wiener filter coefficients by solving Wiener Hopf equation.

Consider input pixel x_kWith Wiener filter output pixel z_k, the wherein output z of Wiener filter_kBy wave filter support { reconstruction pixel y in S}_iComposition, support size is L+1, and weights are c_i.Then Wiener filter function is:

$z_k=\underset{i\&Element;\left\{S\right\}}{\mathrm{\&Sigma;}}{y}_i\&CenterDot;c_i$

Input pixel x_kWith pixel z after Wiener filtering_kBetween residual signals C be defined as:

error_k=z_k-x_k

There is filter tap { c by making_iMean square deviation minimize and optimize Wiener filter:

$c_i=\arg \min E\left[{\mathrm{error}}_k^2\right]$

In order to findMinima, to c_iDifferentiation also derives filter tap by making derivative be equal to zero:

$\frac{\&PartialD;}{\&PartialD;c_i}E\left[{\mathrm{error}}_k^2\right]=2\left(\underset{j\&Element;\left\{S\right\}}{\mathrm{\&Sigma;}}E\right\{\left(y_i\right)\left(y_j\right)\left\}{c}_j\right)-2E\left[\left(y_i\right)\left(x_k\right)\right]=0$ Wherein i=0 ..., L

Note { auto-correlation function of y} and { y} and { cross-correlation function of x} is respectively

r_yy(i)=E[y_k y_k+1] and r_xy(i)=E[x_k y_k+1]

Then above formula can be rewritten as with matrix form:

Thus Wiener filtering coefficient C} can derive as follows with matrix form:

$R_{\mathrm{yy}}\&CenterDot;C=R_{\mathrm{xy}}\&DoubleRightArrow;C=R_{\mathrm{yy}}^{-1}\&CenterDot;R_{\mathrm{xy}}$

5. VS_rec_wiener is added the set of reference frames of this moment target view coded image, and wiener filter coefficients is write Code stream is sent to the coding of three-dimensional video system.

6. in the decoding stage, if the reference frame of current decoded frame is View Synthesis reference frame, use the method identical with decoding end to generate View Synthesis reference frame, then utilizes the Wiener filtering coefficient in code stream that described reference frame is carried out noise reduction filtering, obtains final View Synthesis reference frame.

It can be seen that embodiments provide a kind of introducing enhancing based on Wiener filtering viewpoint from above-mentioned specific embodiment The video coding-decoding method of synthesized reference frame and device, in order to improve compression coding efficiency and the quality of 3 D video.The present invention is real Execute example uses Wiener filter to be filtered optimizing to the reference frame after View Synthesis, improves the quality of virtual visual point image.

It should be noted that in flow chart or at this any process described otherwise above or method describe it is understood that For, represent and include one or more code for the executable instruction of the step that realizes specific logical function or process Module, fragment or part, and the scope of the preferred embodiment of the present invention includes other realization, wherein can not press Order that is shown or that discuss, including according to involved function by basic mode simultaneously or in the opposite order, holds Row function, this should be understood by embodiments of the invention person of ordinary skill in the field.

In the description of this specification, reference term " embodiment ", " some embodiments ", " example ", " concrete example ", Or specific features, structure, material or the feature that the description of " some examples " etc. means to combine this embodiment or example describes comprises In at least one embodiment or example of the present invention.In this manual, the schematic representation to above-mentioned term not necessarily refers to It is identical embodiment or example.And, the specific features of description, structure, material or feature can at any one or Multiple embodiments or example combine in an appropriate manner.

Although above it has been shown and described that embodiments of the invention, it is to be understood that above-described embodiment is exemplary, Being not considered as limiting the invention, those of ordinary skill in the art is in the case of without departing from the principle of the present invention and objective Above-described embodiment can be changed within the scope of the invention, revise, replace and modification.

Claims (6)

1. the coded method of a 3 D video, it is characterised in that including:

S1. depth map and texture maps to reference view encode, and obtain reference view coding and rebuilding depth map D_rec and ginseng Examine viewpoint coding and rebuilding texture maps T_rec；

S2. according to described reference view coding and rebuilding depth map D_rec and reference view coding and rebuilding texture maps T_rec and phase The camera parameter answered, by three-dimension intensity, obtains target view synthesized reference frame VS_rec；

S3. reference view original texture figure T_orig is obtained and as primary signal, by described target view synthesized reference frame VS_rec, as adding the signal after making an uproar, carries out Wiener filtering, obtains optimization aim View Synthesis reference frame VS_rec_wiener, And solve wiener filter coefficients；And

S4. described optimization aim View Synthesis reference frame VS_rec_wiener is added set of reference frames, and by described Wiener filtering Device coefficient write code stream.

2. the coded method of 3 D video as claimed in claim 1, it is characterised in that by solving Wiener Hopf equation meter Calculate described wiener filter coefficients.

3. the coding/decoding method of a 3 D video, it is characterised in that including:

S1. receiving comprises by the code stream obtained according to the coded method of the 3 D video described in any one of claim 1-2, prepare by Frame decoding reference frame image；

S2. judge the type of described reference frame image, if target view synthesized reference frame, then perform S31-S34, if It is independent view reference frame, then performs S4；

S31. from code stream, extract reference view coding and rebuilding depth map D_rec and reference view coding and rebuilding texture maps T_rec；

S32. from code stream, corresponding camera parameter is extracted, in conjunction with described reference view coding and rebuilding depth map D_rec and reference Viewpoint coding and rebuilding texture maps T_rec, by three-dimension intensity, obtains target view synthesized reference frame VS_rec；

S33. from code stream, extract described wiener filter coefficients, described target view synthesized reference frame VS_rec is carried out noise reduction Filtering, obtains final goal View Synthesis reference frame VS_rec_final；With

S34. read the information of final goal View Synthesis reference frame VS_rec_final, complete the decoding process of video image；

S4. directly read the information of described reference frame image, complete the decoding process of video image.

4. the encoder of a 3 D video, it is characterised in that including:

S1. coding and rebuilding module, described reconstruction module, for encoding depth map and the texture maps of reference view, is joined Examine viewpoint coding and rebuilding depth map D_rec and reference view coding and rebuilding texture maps T_rec；

S2. geometric transformation module, described geometric transformation module for according to described reference view coding and rebuilding depth map D_rec and Reference view coding and rebuilding texture maps T_rec and corresponding camera parameter, by three-dimension intensity, obtain target view Synthesized reference frame VS_rec；

S3. Wiener filtering computing module, described Wiener filtering computing module is using reference view original texture figure T_orig as original Signal, using described target view synthesized reference frame VS_rec as adding the signal after making an uproar, carries out Wiener filtering, obtains optimizing mesh Mark View Synthesis reference frame VS_rec_wiener, and solve wiener filter coefficients；And

S4. code stream sending module, described code stream sending module is by described optimization aim View Synthesis reference frame VS_rec_wiener Add set of reference frames, and described wiener filter coefficients is write code stream, send subsequently.

5. the encoder of 3 D video as claimed in claim 4, it is characterised in that in described Wiener filtering computing module, Described wiener filter coefficients is calculated by solving Wiener Hopf equation.

6. the decoder of a 3 D video, it is characterised in that including:

Code stream receiver module, described code stream receiver module is for receiving by according to the 3 D video described in any one of claim 4-5 Encoder transmitted stream, prepare Decoded Reference two field picture frame by frame；

Judge module, described judge module is for judging the type of described reference frame image, if target view synthesized reference Frame, then sequentially enter reconstruction module, geometric transformation module, Wiener filtering computing module and decoding read module, if solely Vertical view reference frame, then be directly entered described decoding read module；

Described reconstruction module, described reconstruction module is for extracting reference view coding and rebuilding depth map D_rec and ginseng from code stream Examine viewpoint coding and rebuilding texture maps T_rec；

Described geometric transformation module, described geometric transformation module is for extracting corresponding camera parameter from code stream, in conjunction with described Reference view coding and rebuilding depth map D_rec and reference view coding and rebuilding texture maps T_rec, by three-dimension intensity, To target view synthesized reference frame VS_rec；

Described Wiener filtering computing module, described Wiener filtering computing module is for extracting described Wiener filter system from code stream Number, carries out noise reduction filtering to described target view synthesized reference frame VS_rec, obtains final goal View Synthesis reference frame VS_rec_final, and it is sent to described decoding read module as the reference frame image after updating；

Described decoding read module, for reading the information of described reference frame image, completes the decoding process of video image.